Intel Patents Redundant Cores In a Many-Core Processor

[climber]

Sounds like RAID, except RAIP, Redundant Arrangement of Interconnected Processors, of course the word has an unfortunate pronunciation

 

[ElectroGoofy]

[citation][nom]JacekRing[/nom]At least they didn't patent a square processor with rounded corners....[/citation]
My guess is that Apple beat them to it.

 

[buzznut]

This sounds like how the brain works. There is redundancy built into the human brain, in fact a whole bunch of redundancy. For instance, retrieving a memory can fail for many reasons but the brain can choose a different pathway to get to the memory when this happens. This is why the brain has so many damn connections, to accommodate all this redundancy. The human brain is very inefficient, and yet is fascinating in what it can accomplish due to all these millions of connections.
Conceivably with enough cores and connections, you could get a CPU to be capable of creative thought.

 

[spookyman]

[citation][nom]outlw6669[/nom]This is a terrible idea.Why not ship the CPU with all cores active and give it a 'soft fail' feature for failing cores?By 'soft fail' I mean that a failing core could be dynamically deactivated while allowing all other cores to function normally.This would allow you to have higher initial performance and give you uninterrupted computing in the case of a core failure.[/citation]

Normally Space Equipment tends to not use Intel components there.

 

[coldmast]

"temperature thus has a direct bearing on core MTTF [mean time to failure] and many-core reliability"
That's why I don't overclock and have a non-stock HS/F and also why I have so many darn case fans. Also, I patent the ability to sporadically alter (or flux) which cores are on and receiving instructions (like a minigun rotates barrels to keep from melting).

 

[coldmast]

[citation][nom]Lyden[/nom]Um... when was the last time you had a core fail on you? Me? never.[/citation]
I had a Pentium 4 (2.4 Ghz) fail on me, I had system errors (just like with hard drive problems) w/o thermal tripping and one day it just wouldn't boot (like with failed ram). I cross tested the components with a different computer, and I even thought it was motherboard, after that was switched it turned out to be the processor - so it can happen.

 

[haschid]

I thought PS3 Cell processor already had redundant SPE's.

 

[capt_taco]

Wow, they really will give out a patent for anything.

I need to file my patent for fire, and become a trillionaire.

 

[hetneo]

[citation][nom]Zingam[/nom]And that shit is patentable, right?[/citation]
Things like this shouldn't be patentable. Hell if they get patent for this I'm submitting patent application for having two wrenches of same size for case that one fails due to effects of lateral torque force, and I will also draw pretty diagram with allocation of 1st wrench to active status and allocation of 2nd wrench to spare status, and I will also put on it a step for inspecting spare wrench for cracks.

Damn, I'm so smart I could work for Intel

 

[elbert]

AMD has already been doing this so possible the patent is waste of time. Most AMD motherboards will allow you to disable any given core. IE in the 960t you can unlock one of the disabled cores for any active. Not sure if the automatic part would mean enough to all the patent to hold up in court.

Not sure this would even be different enough from underclocking hot cores to allow a patent.

 

[aznguy0028]

[citation][nom]buzznut[/nom]This sounds like how the brain works. There is redundancy built into the human brain, in fact a whole bunch of redundancy. For instance, retrieving a memory can fail for many reasons but the brain can choose a different pathway to get to the memory when this happens. This is why the brain has so many damn connections, to accommodate all this redundancy. The human brain is very inefficient, and yet is fascinating in what it can accomplish due to all these millions of connections. Conceivably with enough cores and connections, you could get a CPU to be capable of creative thought.[/citation]
Nope, you are completely wrong. The human brain is actually very efficient. The period of time you are talking about is within the first few months of life to the first year. At that time, our brain actually has more neurons than we do as adults. However, as time goes by, the brain lateralizes and certain activities (vision, hearing, etc) become much more focused in certain areas than others. This whole process is called "synaptic pruning", where all the synapses that were unused goes away, hence making the brain MUCH MORE efficient. The reason why we have so many synapses and neurons when we were young is because if an injury or accident happens to our brain, it can recover way more easily than as adults. From an evolutionary point of view, this makes sense as it help promotes the survival of our species. In human development terms, go read up "plasticity", it would enlighten you on the topic.

 

[Cazalan]

[citation][nom]t_wilson[/nom]Yet another example of just how ridiculous the patent system is.[/citation]

Agreed. Redundancy is not a novel idea. Having a fail over is not a novel idea.
The ACPI specification (1996) already included processor states from mechanically off, to sleeping, to on.

Having a real time in-system mechanism to determine a failing core might be novel. Figure 140

 

[QEFX]

Might be wrong, and I usually am, but isn't this type of thing used in mainframes? Active/inactive cores that can be activated for a variety of reasons for a period of time and support for redundant parts for high availability.

 

[twelch82]

When we get more cores, the software will start to catch up. I do a lot of lock-free multithreaded programming (the easier and more performance oriented kind), and you typically have to write the code differently to support multiple threads. There is an upfront memory and performance cost, but once it is paid, scaling across multiple threads is typically close to linear.

Problem is, if half of the people are running the code on machines that only have two threads, by the time that overhead is factored in, it may not be worth it. If everyone has a processor with lots of cores, it will be hard to leave that extra performance on the table.

 

[hawkwindeb]

[citation][nom]outlw6669[/nom]This is a terrible idea.Why not ship the CPU with all cores active and give it a 'soft fail' feature for failing cores?By 'soft fail' I mean that a failing core could be dynamically deactivated while allowing all other cores to function normally.This would allow you to have higher initial performance and give you uninterrupted computing in the case of a core failure.[/citation]

Putting aside the issue already mentioned about not having all cores active due to heat & power that might not be desirable for a given use case, there exist today and since 2005 a processor that can have one or more of its cores "proactively off-lined" while the operating system and its applications/service it's running are still live and processing client requests, etc. It works by monitoring soft errors being corrected in HW up to a given threshold and once passed that threshold, it predicts that the item is going to fail and then off-lines the item before the system crashes due to a failure of that item. High-level description but working for years and even enhanced to off line individual threads within cores. The processor, SPARC T-Series including threads and other items within the processor and SPARC64 VI through latest VII+ (AKA M-Series) with Solaris 10 and Solaris 11. Can off line other components but this is just a processor core discussion, so nuff said. Also running Solaris 10 or 11 on some of the latest x86 processors provides a subset of some of the SPARC-Solaris combo off-lining capabilities.

 

[law shay]

[citation][nom]xX_PEMDAS_Xx[/nom]Who would want a CPU that gets slower over time?? I want a bulldozer, its cores come stable for less money.[/citation]
Bulldozer is good, no doubt.
But are you nuts ? Do you know what the news article is saying about ? Do you understand a bit about this patent ?
It has nothing to do with slowing down the cores. It is talking a different thing.

 

[law shay]

[citation][nom]capt_taco[/nom]Wow, they really will give out a patent for anything.I need to file my patent for fire, and become a trillionaire.[/citation]

I'm afraid you are talking about my patent-pending innovation called:

Heat-and-Light-producing plasma obtained through burning any mateirals that has a mass in either a right mix oF gases or In absent theREof" or in accronym, "FIRE".

I'm gonna sue you.

 

[eddieroolz]

Its a nice idea in theory, but the thing is CPUs are expected to not fail...

 

[ProDigit10]

Instead of just using better ventilating systems in their computers and pc's!

it costs about $20, to create a plastic tube, that will channel airflow from a fan past the sink to an exiting fan (2 fans for failsafe), and have an optimal airflow!